To install the package, use the following command:
devtools::install_github("davidruegamer/deepregression")
Note that the installation requires additional packages (see below) and their installation is currently forced by deepregression
.
The requirements are given in the DESCRIPTION
. If you load the package manually using devtools::load_all
, make sure the following packages are availabe:
- Matrix
- dplyr
- keras
- mgcv
- reticulate
- tensorflow
- tfprobability
If you set up a Python environment for the first time, install reticulate
and run the check_and_install
function from the deepregression
package. This tries to install miniconda, TF 2.1, TFP 0.9 and keras 2.4.3, which seems to be the most reliable setup for deepregression
at the moment.
Until published, please cite the following preprint:
@article{rugamer2020unifying,
title={Semi-Structured Deep Distributional Regression: Combining Structured Additive Models and Deep Learning},
author={R{\"u}gamer, David and Kolb, Chris and Klein, Nadja},
journal={arXiv preprint arXiv:2002.05777},
year={2020}
}
See the tutorial for a detailed introduction.
A Python version of the package is available here.
The following works are based on the ideas implemented in this package: